Topic: EWBF's CUDA Zcash miner - page 254. (Read 2164327 times)

Hi guys, what is best setup for Gigabyte G1 Gaming gtx 1060 3gb, Samsung memory? Currently I getting about 260 with +160 (core), +550 (mem), powerlimit 68. Last NV drivers 376.48.

As I said overclocking is impossible because even with default clocks miner crashes, I actually have to underclock by -200+ to get it working but unstable, -280 gives stable enough results.

I had quite a bit of instability with 375.x try upgrading to 376.x drivers. Also from what i read around msi cards arent very reputable in OC stability. But try it at least

Extremely strange card behaviour...

This is default settings without being touched.OC is impossible for GPU core.

Lowering the core below -250 seems to do the trick.
Memory +200 without errors. If I go above +200 strange noise appears.

Any suggestions/ideas much appreciated!

EDIT: This is supposed to be MSI Armor GTX 1070 8GB but is recognized as Armor OC version by SN also.
Found latest bios on MSI site(supposed to fix Micron memory crashing) and upgraded it, but the error still remains...

ewbf why am i getting less hashrate using 0.0.8?

970 gets a bit less hashrate with 0.0.8 than 0.0.7. Does anyone else have this problem with maxwell cards?
260 sol/s with 0.0.7 and 250 sol/s with 0.0.8

Same OC as usual 50% tdp +135/600 20% vCore, i think i used something like 16x75 and played a bit with others options to make it go a bit faster. Tried quite a while ago, don't remember the exact settings. By default with no special switch it was giving me something around 4kh, played a bit to grab 150/200 more.

it doesn't work for cuda 2.1 cards,giving 0 sols,card is geforce 710m

what setting give you 700 hash for a single 1070? the best one i found is 12x60 intensity but give me only 600 at best

Have you tried the cuda 8 version of tsiv ccminer by KlausT ? it's far better than what basic tsiv can give you (i do 4200h XMR with 6x1070 rig but i can't max it out using 100w less @50tdp than zcash), zcash is still more interesting, but it could become a good contender when zcash drop , i agree that XMR gpu mining could/should be improved. Interesting query. I wonder why claymore isn't working more with nvidia...

I wouldn't be asking here if I thought it worked well. Cryptonote hasn't really been improved for some time now. Just compiling for CUDA8 doesn't really make much of a difference.

Claymore has always been AMD biased. From what I read in his thread with Ethereum he was paid to not support Nvidia for about six months before it expired and he added Nvidia support. He's a very capable program, he usually bases decisions like this off of business.

Have you tried the cuda 8 version of tsiv ccminer by KlausT ? it's far better than what basic tsiv can give you (i do 4200h XMR with 6x1070 rig but i can't max it out using 100w less @50tdp than zcash), zcash is still more interesting, but it could become a good contender when zcash drop , i agree that XMR gpu mining could/should be improved. Interesting query. I wonder why claymore isn't working more with nvidia...


For my personnal knowledge what is the issue with coinotron pool, that would need special support ?

Thnaks a lot.
I compiled it (I am with Linux and no release for Linux) and try it : unfornately, it is a lot slower han my curently miner (TSIV) for my 750s :-(.

right, thats what i roughly had in mind, thanks for confirming!  
sec.suprnova.cc also allows choosing your own diff btw.

i also remember that it was recommended to ideally have each node send 4-10 shares per minute to the server in order to reduce load on the backend which in turn should benefit the whole pool.  

now what i wanted to better understand is what happens in cases where my nodes often get new work assigned without having sent any shares since the last work-package was assigned. are the computations done in this timeframe lost (to my disadvantage) and would thereby sending rather 10-20 (accepted)shares per minute lead to a higher accepted number of shares?
according to my (unproven) theory, i would loose less work if shares are sent more frequently. of course this has not to be the case every 3 seconds or so but i've observed quite a few cases where new workloads were sent only 10-20 seconds apart from another...

thanks for enlightening me  

https://github.com/KlausT/ccminer-cryptonight/releases

There you go, friend

I get 225-230 sol/s @ 65W (55% power limit). Core +185 / memory +240.
That's that, can't compare 1060 with 1070/80.

How about you go and read the previous pages, several peoples posted their settings... alot of them....

#!/bin/bash
sudo X :0 &
export DISPLAY=:0
CLOCK=150
MEM=700
PL=110
FANSPEED=70
CMD='/usr/bin/nvidia-settings'

for i in {0..5} # for 6 cards
  do
sudo nvidia-smi -i ${i} -pm 0
sudo nvidia-smi -i ${i} -pl {$PL}
sudo ${CMD} -a [gpu:${i}]/GPUPowerMizerMode=1
sudo ${CMD} -a [gpu:${i}]/GPUFanControlState=1
sudo ${CMD} -a [fan:${i}]/GPUTargetFanSpeed=${FANSPEED}

  for x in {3..3}
      do
      sudo  ${CMD} -a [gpu:${i}]/GPUGraphicsClockOffset[${x}]=${CLOCK}
      sudo  ${CMD} -a [gpu:${i}]/GPUMemoryTransferRateOffset[${x}]=${MEM}
    done
done

#!/bin/bash
# Paths to the utilities we will need
export DISPLAY=:0
SMI='/usr/bin/nvidia-smi'
SET='/usr/bin/nvidia-settings'

# Determine major driver version
VER=`awk '/NVIDIA/ {print $8}' /proc/driver/nvidia/version | cut -d . -f 1`

# Drivers from 285.x.y on allow persistence mode setting
if [ ${VER} -lt 285 ]
then
    echo "Error: Current driver version is ${VER}. Driver version must be greater than 285."; exit 1;
fi

# Read a numerical command line arg between 40 and 100
if [ "$1" -eq "$1" ] 2>/dev/null && [ "0$1" -ge "40" ]  && [ "0$1" -le "100" ]
then
    speed=$1   # set speed

    echo "Setting fan to $speed%."

    # how many GPU's are in the system?
    NUMGPU="$(nvidia-smi -L | wc -l)"

    # loop through each GPU and individually set fan speed
    n=0
    while [  $n -lt  $NUMGPU ];
    do
        # start an x session, and call nvidia-settings to enable fan control and set speed
        ${SET} -a [fan:${n}]/GPUTargetFanSpeed=$speed
        let n=n+1
    done

    echo "Complete"; exit 0;

elif [ "x$1" = "xstop" ]
then
    echo "Enabling default auto fan control."

    # how many GPU's are in the system?
    NUMGPU="$(nvidia-smi -L | wc -l)"

    # loop through each GPU and individually set fan speed
    n=0
    while [  $n -lt  $NUMGPU ];
    do
        # start an x session, and call nvidia-settings to enable fan control and set speed
        ${SET} -a [gpu:${n}]/GPUFanControlState=0
        let n=n+1
    done

    echo "Complete"; exit 0;

else
    echo "Error: Please pick a fan speed between 40 and 100, or stop."; exit 1;
fi

#!/bin/bash
GPU='0 1 2 3 4 5'
STRATUM_SERVER="eu1-zcash.flypool.org"
STRATUM_PORT="3333"
POOL_WORKER="Your_Wallet"
POOL_PASS="x"
# Comment next lines if you experience any issues
echo 'Doing some stuff...maybe not related to NVIDIA...'
export GPU_MAX_HEAP_SIZE=100
export GPU_USE_SYNC_OBJECTS=1
export GPU_MAX_ALLOC_PERCENT=100
export GPU_SINGLE_ALLOC_PERCENT=100
sleep 1
echo 'Starting EWBF's CUDA Zcash miner, use "screen -r cuda_miner" to resume...'
killall -TERM miner >/dev/null 2>&1
sleep 2
/usr/bin/screen -dmS cuda_miner ./miner --server ${STRATUM_SERVER} --port ${STRATUM_PORT} --user ${POOL_WORKER} --pass ${POOL_PASS} --cuda_devices ${GPU}
exit 0